Our study shows that the elderly spent 65% of their daily time being sedentary and 35% being active, of which only 2% was MVPA. These results are similar to those reported from other elderly samples [19–21]. Duration of activity bouts decreased with increasing intensity level, i.e. substantially more minutes are continuously spent sedentary than in light activity or MVPA. Other studies which examined the patterns of elderly people in different activity levels are rare. Lord et al.  examined the G of sedentary activity and walking in elderly people and presented similar results to ours: sedentary bouts tended to be longer than walking bouts. In contrast, Donaire-Gonalez and colleagues  did not find differences concerning the duration of bouts in MVPA and overall PA. However, in their study, the duration of bouts was expressed as the median bout length which may be a less significant measure than the G value itself.
Subjects in the frequently active group were more active than others at all levels of PA, spending more time in both light activity and MVPA than the comparison groups and less time sedentary. PA patterns differed by degree of activity (‘rare’, ‘average’, and ‘frequent’) among all intensity levels. Findings were most clear with regard to the time in MVPA as well as GMVPA: the more active a person was, the larger was the proportion of long bouts in light and MVPA and the larger was the proportion of short bouts in the sedentary level.
To our knowledge, the work of Chastin et al.  was the only other publication that compared PA patterns of more and less active people. They focused on patterns of sedentary activity and found that the sedentary time of less active subjects was composed of longer rest periods, which is in accordance with our findings.
PA time was correlated with PA patterns: an increased proportion of long bouts (higher G values) was positively correlated with the accumulated PA time in the respective intensity level. This effect was most significant for MVPA. Comparisons between the different activity groups support this finding. Accordingly, a higher proportion of long bouts in light PA as well as in MVPA may be beneficial to increase the overall activity time in rarely active people. In contrast, given a certain activity level, the proportion of long bouts in MVPA must be increased in order to increase the overall activity level and related health benefits , as patterns of light PA do not differ significantly between the average and frequent group (see Additional file 2: Table S2).
However, these recommendations may need to be adjusted for each individual, particularly for people in the rare group. Exercise capacity and specific health conditions [2, 25] may limit the ability to perform PA. In our sample, rarely active people typically had reduced FEC, poor lung function, higher prevalence of multimorbidity, and higher prevalence of disability. However, many disabled or multimorbid persons in our sample achieved higher levels of PA than people without disability or multimorbidity. Thus, disability and multimorbidity do not necessarily limit PA.
We tested associations between PA times and G values with age, gender, and BMI within each intensity level, in order to examine whether and how particular intensity levels and patterns differ in specific groups. Older participants, obese persons and women seem to be particularly prone to a sedentary lifestyle. The reduced average activity of obese subjects is particularly associated with decreased time in MVPA, whereas the reduced average activity of elderly participants is due to a decreased time in both light PA and MVPA.
Examinations of the relationship between PA and gender are inconsistent. Gardner and colleagues , for example, demonstrated that women with intermittent claudication aged 65 ambulate slower than men. Jakicic et al.  objectively measured the MVPA patterns of 59-year-old overweight and obese individuals with type 2 diabetes mellitus, and found that men have a larger amount of bouts ≥10 minutes in that level than women. Those findings agree with ours, whereas their results of BMI and age are contrary: they found no associations between MVPA bouts and age or BMI . In line with our results, other studies found that the proportion of long bouts (8–10 min/day) of MVPA declines with increasing BMI and advancing age [27, 28]. Lastly, a study with subjects aged 70–88 identified younger age and lower BMI as significant predictors of walking. There was no correlation between PA and gender .
Younger age, lower BMI, male sex, better lung function, absence of multimorbidity, more time and longer bouts (higher G values) in the MVPA level, emerged as significant predictors of exercise capacity: they explained 56% of the total variance in FEC. It is important to note that GMVPA accounted for 27% of the variance, by far the largest single predictor. This finding indicates that in addition to the well-known relationship between FEC and PA in terms of duration and intensity, there is also one in terms of patterns. Correlations between FEC and PA characteristics (such as intensities or patterns) have been examined and evidenced before [29–31]. However, these studies either used univariate analysis or failed to consider detailed information about PA (like intensities and patterns). Hernandes et al. , for example, demonstrated that the intensity of movement correlates with 6MWD in healthy elderly individuals (r = 0.49; p < 0.01) and that walking time is positively associated with FEC in COPD patients (r = 0.42; p < 0.01). However, no information about further predictors of FEC was shown. Moreover, no advice about the level of intensity is given.
Our findings underline the current PA guidelines for older adults [2, 25], which imply that activity should be at least moderate intensity. Moreover, our results support the fact that a higher proportion of longer bouts predict FEC better than a higher proportion of shorter bouts, with potential greater effect on health .
The present study is the first one that examined PA patterns in terms of G among three different intensity levels (sedentary PA, light PA, and MVPA) and thus presents detailed and differentiated information about activity patterns of different intensities in elderly people. We identified associations between times and patterns of PA in different intensity levels and examined the relationship regarding FEC. However, due to the cross-sectional study design the direction of causality of the examined associations cannot be assured.
Recognized limitations of accelerometers include their inability to detect non-walking activity such as resistance training or cycling. Thus they are likely to underestimate such activities  and related bout lengths. Another limitation of this study is the questionable validity of the cut-points applied to classify activities into intensity levels. No general standard for transforming activity data into different intensity levels exists  although there are many validation and calibration studies. We chose the algorithm by Freedson et al.  because it is the most often used validated algorithm for ActiGraph sensors  and therefore has the highest potential to provide comparable data. Since the calibration study of Copeland and Esliger  was been performed specifically with older adults, we also present PA variables based on these cut points in order to increase the comparability of this novel method for the future. As expected, differences for time spent in MVPA are detectable, particularly among individuals who are infrequently active. Direct comparison of PA variables would enable systematic measurement of how choice of cut-points influences prediction of PA levels and patterns. We consider this a very interesting scientific issue and plan to discuss it in a separate paper.
In conclusion, both time spent in MVPA and GMVPA emerged as important predictors for functional exercise capacity. Time in MVPA can most profitably be increased by increasing the proportion of long bouts which enhances activity levels and meets recommendations for PA while simultaneously increasing G. These recommendations can be followed by most older adults, but those with functional or health-related limitations may need to adjust accordingly. In rarely active people (commonly characterized by higher age, higher BMI, reduced functional exercise capacity, worse lung function, and multimorbidity and disability) even light activity, i.e. a higher proportion of long bouts in light PA, may increase average activity levels.